VDURA launches Sentinel to auto-diagnose storage faults

VDURA has launched Sentinel for the VDURA Data Platform V12. The support tool is included with the latest platform release.

Sentinel continuously monitors the storage platform and can open a service request with VDURA support when it detects a fault pattern. The ticket can include diagnostic information, a proposed fix and, where required, trigger a parts dispatch through the support service.

The product is aimed at operators of GPU clusters and cloud environments, where storage failures can interrupt training jobs and leave expensive computing resources idle. Sentinel analyses health information from customer systems against patterns drawn from thousands of field deployments.

Instead of waiting for an operator to spot a problem, collect logs and file a support case, the system is designed to start that process automatically. VDURA says the first support engineer reviewing the case sees an existing diagnosis rather than starting from scratch.

How it works

Sentinel runs natively on the V12 platform and uses the same published application programming interfaces as the rest of the software. The telemetry available to VDURA's support team is also available to customers through their own dashboards and observability tools.

There is no separate vendor-only access layer. Customers choose which subset of operational data to share with VDURA support through the Sentinel cloud service.

Only four categories of metadata can be transmitted: fleet health and capacity metrics, node inventory, alerts and deep-dive diagnostics. File contents, credentials, user information, workload details and logs are not sent.

Security controls

According to VDURA, the collector runs inside the customer environment with read-only credentials and makes outbound HTTPS connections only. The design means there is no inbound access and no ability for the tool to alter configuration or access stored data.

Each of the four telemetry categories can be disabled independently, or data sharing can be turned off completely, without affecting the local dashboard or the underlying storage system. Customers can also inspect the exact contents of each outbound payload before transmission.

VDURA says the collector is published as dependency-free, readable Python code for review by customer security teams. The approach appears intended to address concerns among infrastructure operators that do not want opaque monitoring agents inside tightly controlled environments.

The launch comes as suppliers to AI infrastructure operators face pressure to reduce downtime in systems built around large GPU clusters. In these environments, any interruption in storage can have wider operational consequences, because compute resources may sit idle while support teams investigate faults.

Sentinel is included in the V12 release, alongside a local fleet health dashboard and access to the underlying telemetry through APIs. Automated service requests and parts dispatch are available through VDURACare Premier, the company's support offering for hardware, software and round-the-clock response.

Erik Salo, Senior Vice President of Marketing and Operations at VDURA, said the launch is intended to shorten the time between fault detection and resolution by using information gathered across the company's installed base.

"Every storage failure a customer sees is one we have almost certainly already seen and resolved somewhere else in the fleet," said Erik Salo, Senior Vice President of Marketing and Operations at VDURA.

"Sentinel puts that fleet-wide knowledge to work automatically so the first person to look at a ticket is looking at a diagnosis instead of a blank page. For a GPU cloud, the gap between fault and fix is measured in idle accelerators and tenant SLAs. Closing that gap is what proactive support should mean," Salo said.